Methods for automated habitat mapping &potential relationship with strategic monitoringSander Mücher1, Rob Jongman1, Bob Bunce2, Henk Kramer11) Wageningen Environmental Research, 2) Estonian University of Life Sciences TartuCost Action HARMONIOUS, Workshop standardization of procedures in usingUAS for environmental monitoring6th of November 2019, University of CoimbraContact: [email protected]

WENR remote sensing domain & trends Land monitoring Monitoring vegetation for agriculture &nature Using satellites & drones Big Data, Cloud computing, Machine learning

WUR - Unmanned Aerial Remote SensingFacility (UARSF) since 2012ROC certified

BackgroundWhy we need new strategic European habitatmonitoring ? Habitats are important indicators of biodiversity intheir own right, and have a strong link with species Habitats are used a lot in reporting on biodiversity For reporting consistent mapping and monitoring habitatextent and change is important Remote Sensing techniques are becoming an importanttool for this.

Monitoring objectives Collect European habitatinformation consistently fromeach country, not only inprotected sides but also in thewide countryside (unbiased)across all environmental zones Provide consistent Europeanstatistics for habitats Provide strict protocols forhabitat mapping & monitoring Support existing habitat &vegetation monitoring schemes

Requirements for unbiased Europeanhabitat monitoring Stratified random samples of 1km2 across EnvironmentalZones in Europe, such as used in UK Countryside Survey Wall-wall habitat mapping in km2 , including areal, linear, pointhabitats Samples should include: 1) Natura 2000 sites; 2) twinned samplesoutside; 3) LTER sites Habitats of restricted distribution which will have to be targetedUse European Habitat typology (EUNIS/General Habitat Categories)Plant life forms centralDominant species per life formRecording of environ. qualifiers optionalSurveyors must be trained

Field computer help standardized in-situ habitat mappingDownloadable from:

European habitat modelling spatial of allEUNIS habitat types to assist sampling

Many spaceborne EO approaches alreadyexist for automated habitat mappingVHRSEODHAMmultitemporaldecision treesLucas et al, 2015. The Earth Observation Data for Habitat Monitoring (EODHaM) System.International Journal of Applied Earth Observation and Geoinformation 37 (2015) 17–28.Mücher, et al., 2015. Synergy of Airborne LiDAR and Worldview-2 satellite imagery for land coverand habitat mapping: a BIO SOS-EODHaM case study for the Netherlands. International Journal ofAppliedEarth Observation and Geoinformation 37 (2015) 48–55.

Rule based versus Machine learningRule based classification (RB)Accuracy84.1 %Random Forest classification (RF)Accuracy86.4 %10Mücher, et al., 2019. Journal of Earth Sciences & Environmental Studies, VOLUME: 4 ISSUE: 1, pp 502-505. DOI:10.25177/JESES.4.1.2

Vegetation structure monitoring possible for entireNetherlands (at scale 1:5000) once in the 6years based on free VHRS and national LiDAR dataGeoEye-1 22 June 2009 Veg cover types 2009 Vegetation height 2009from National LiDAR dataBareCoveredVegetatedWV-2 27 August 2014Veg cover types 2014 Vegetation height 2014

Why use of UAVs ?UAV hyperspectral vegetationclassification at 2 cm resolution

Advantages UAVs over VHR satelliteimagery UAV imagery has a much finer spatial resolution than VHR satelliteimagery enables to map small areal and linear habitats ( vegetationmapping) UAV imagery are easier to use for vegetation surveyors that are used toaerial photography UAV imagery can nowadays easily cover square km samples at therequired time periods of the year (up-to-date at right moments)without cloud cover Combining multispectral and LiDAR camera’s possible on UAV platforms(national LiDAR data is not up-to-date). So up-to-date and accurate habitat maps and changes possible withdrones across sample sites. Remotely classified habitat maps and associated changes as basis foradditional field surveys

Standardization UAV data for habitatmonitoring Choice platform and camera types should be flexible dueto user requirements but imagery should be consistentin time due to monitoring Protocols for processing imagery using photographicprinciples, including positional accuracy Flight execution should follow strict rules and shouldbe repeated consistently over time. Rules for multi-time recording versus weather conditionsneeds smart decision making Sander Mücher, Wageningen Environmental Research (Alterra)Tel: 31 (0) 317 481607, email: [email protected]

Requirements for unbiased European habitat monitoring Stratified random samples of 1km2 across Environmental Zones in Europe, such as used in UK Countryside Survey Wall-wall habitat mapping in km2 , including areal, linear, point habitats Samples should include: 1) Natura 2000 sites; 2) twinned samples